Oil pipeline leak signal image recognition based on improved data field theory

Abstract

To maximize use of the valid features of time series signal images, and to detect oil pipeline leaks rapidly, accurately and reliability, the theory of data field is used for its advantages in clustering and singular value recognition. First, the feasibility of using acoustic wave signals for leak detection is demonstrated. Then, the semi-hard semi-soft thresholding function is used for de-noising. This method not only reduces the constant deviation in wavelet-based soft thresholding and hard thresholding, but also preserves the original features of signals and makes the de-noised signals smooth. Finally, the application of data field theory for leak detection and localization is analyzed and an improved algorithm based on data field theory is proposed. And the accuracy and universality of the proposed algorithm are verified through experiments. It is found that the adjusting parameters, influence factors and the width of the sliding window only affect the amplitude of the potential curve. That is, the localization of leak signals is not affected. Research shows that the proposed algorithm is a simple and effective new method for pipeline leak detection and localization besides correlation algorithm.